xref: /openbmc/linux/fs/nilfs2/the_nilfs.h (revision 65417d9f)
1 /*
2  * the_nilfs.h - the_nilfs shared structure.
3  *
4  * Copyright (C) 2005-2008 Nippon Telegraph and Telephone Corporation.
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License as published by
8  * the Free Software Foundation; either version 2 of the License, or
9  * (at your option) any later version.
10  *
11  * This program is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  * GNU General Public License for more details.
15  *
16  * Written by Ryusuke Konishi.
17  *
18  */
19 
20 #ifndef _THE_NILFS_H
21 #define _THE_NILFS_H
22 
23 #include <linux/types.h>
24 #include <linux/buffer_head.h>
25 #include <linux/rbtree.h>
26 #include <linux/fs.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/slab.h>
30 #include <linux/refcount.h>
31 
32 struct nilfs_sc_info;
33 struct nilfs_sysfs_dev_subgroups;
34 
35 /* the_nilfs struct */
36 enum {
37 	THE_NILFS_INIT = 0,     /* Information from super_block is set */
38 	THE_NILFS_DISCONTINUED,	/* 'next' pointer chain has broken */
39 	THE_NILFS_GC_RUNNING,	/* gc process is running */
40 	THE_NILFS_SB_DIRTY,	/* super block is dirty */
41 };
42 
43 /**
44  * struct the_nilfs - struct to supervise multiple nilfs mount points
45  * @ns_flags: flags
46  * @ns_flushed_device: flag indicating if all volatile data was flushed
47  * @ns_sb: back pointer to super block instance
48  * @ns_bdev: block device
49  * @ns_sem: semaphore for shared states
50  * @ns_snapshot_mount_mutex: mutex to protect snapshot mounts
51  * @ns_sbh: buffer heads of on-disk super blocks
52  * @ns_sbp: pointers to super block data
53  * @ns_sbwtime: previous write time of super block
54  * @ns_sbwcount: write count of super block
55  * @ns_sbsize: size of valid data in super block
56  * @ns_mount_state: file system state
57  * @ns_sb_update_freq: interval of periodical update of superblocks (in seconds)
58  * @ns_seg_seq: segment sequence counter
59  * @ns_segnum: index number of the latest full segment.
60  * @ns_nextnum: index number of the full segment index to be used next
61  * @ns_pseg_offset: offset of next partial segment in the current full segment
62  * @ns_cno: next checkpoint number
63  * @ns_ctime: write time of the last segment
64  * @ns_nongc_ctime: write time of the last segment not for cleaner operation
65  * @ns_ndirtyblks: Number of dirty data blocks
66  * @ns_last_segment_lock: lock protecting fields for the latest segment
67  * @ns_last_pseg: start block number of the latest segment
68  * @ns_last_seq: sequence value of the latest segment
69  * @ns_last_cno: checkpoint number of the latest segment
70  * @ns_prot_seq: least sequence number of segments which must not be reclaimed
71  * @ns_prev_seq: base sequence number used to decide if advance log cursor
72  * @ns_writer: log writer
73  * @ns_segctor_sem: semaphore protecting log write
74  * @ns_dat: DAT file inode
75  * @ns_cpfile: checkpoint file inode
76  * @ns_sufile: segusage file inode
77  * @ns_cptree: rb-tree of all mounted checkpoints (nilfs_root)
78  * @ns_cptree_lock: lock protecting @ns_cptree
79  * @ns_dirty_files: list of dirty files
80  * @ns_inode_lock: lock protecting @ns_dirty_files
81  * @ns_gc_inodes: dummy inodes to keep live blocks
82  * @ns_next_generation: next generation number for inodes
83  * @ns_next_gen_lock: lock protecting @ns_next_generation
84  * @ns_mount_opt: mount options
85  * @ns_resuid: uid for reserved blocks
86  * @ns_resgid: gid for reserved blocks
87  * @ns_interval: checkpoint creation interval
88  * @ns_watermark: watermark for the number of dirty buffers
89  * @ns_blocksize_bits: bit length of block size
90  * @ns_blocksize: block size
91  * @ns_nsegments: number of segments in filesystem
92  * @ns_blocks_per_segment: number of blocks per segment
93  * @ns_r_segments_percentage: reserved segments percentage
94  * @ns_nrsvsegs: number of reserved segments
95  * @ns_first_data_block: block number of first data block
96  * @ns_inode_size: size of on-disk inode
97  * @ns_first_ino: first not-special inode number
98  * @ns_crc_seed: seed value of CRC32 calculation
99  * @ns_dev_kobj: /sys/fs/<nilfs>/<device>
100  * @ns_dev_kobj_unregister: completion state
101  * @ns_dev_subgroups: <device> subgroups pointer
102  */
103 struct the_nilfs {
104 	unsigned long		ns_flags;
105 	int			ns_flushed_device;
106 
107 	struct super_block     *ns_sb;
108 	struct block_device    *ns_bdev;
109 	struct rw_semaphore	ns_sem;
110 	struct mutex		ns_snapshot_mount_mutex;
111 
112 	/*
113 	 * used for
114 	 * - loading the latest checkpoint exclusively.
115 	 * - allocating a new full segment.
116 	 */
117 	struct buffer_head     *ns_sbh[2];
118 	struct nilfs_super_block *ns_sbp[2];
119 	time_t			ns_sbwtime;
120 	unsigned int		ns_sbwcount;
121 	unsigned int		ns_sbsize;
122 	unsigned int		ns_mount_state;
123 	unsigned int		ns_sb_update_freq;
124 
125 	/*
126 	 * The following fields are updated by a writable FS-instance.
127 	 * These fields are protected by ns_segctor_sem outside load_nilfs().
128 	 */
129 	u64			ns_seg_seq;
130 	__u64			ns_segnum;
131 	__u64			ns_nextnum;
132 	unsigned long		ns_pseg_offset;
133 	__u64			ns_cno;
134 	time_t			ns_ctime;
135 	time_t			ns_nongc_ctime;
136 	atomic_t		ns_ndirtyblks;
137 
138 	/*
139 	 * The following fields hold information on the latest partial segment
140 	 * written to disk with a super root.  These fields are protected by
141 	 * ns_last_segment_lock.
142 	 */
143 	spinlock_t		ns_last_segment_lock;
144 	sector_t		ns_last_pseg;
145 	u64			ns_last_seq;
146 	__u64			ns_last_cno;
147 	u64			ns_prot_seq;
148 	u64			ns_prev_seq;
149 
150 	struct nilfs_sc_info   *ns_writer;
151 	struct rw_semaphore	ns_segctor_sem;
152 
153 	/*
154 	 * Following fields are lock free except for the period before
155 	 * the_nilfs is initialized.
156 	 */
157 	struct inode	       *ns_dat;
158 	struct inode	       *ns_cpfile;
159 	struct inode	       *ns_sufile;
160 
161 	/* Checkpoint tree */
162 	struct rb_root		ns_cptree;
163 	spinlock_t		ns_cptree_lock;
164 
165 	/* Dirty inode list */
166 	struct list_head	ns_dirty_files;
167 	spinlock_t		ns_inode_lock;
168 
169 	/* GC inode list */
170 	struct list_head	ns_gc_inodes;
171 
172 	/* Inode allocator */
173 	u32			ns_next_generation;
174 	spinlock_t		ns_next_gen_lock;
175 
176 	/* Mount options */
177 	unsigned long		ns_mount_opt;
178 
179 	uid_t			ns_resuid;
180 	gid_t			ns_resgid;
181 	unsigned long		ns_interval;
182 	unsigned long		ns_watermark;
183 
184 	/* Disk layout information (static) */
185 	unsigned int		ns_blocksize_bits;
186 	unsigned int		ns_blocksize;
187 	unsigned long		ns_nsegments;
188 	unsigned long		ns_blocks_per_segment;
189 	unsigned long		ns_r_segments_percentage;
190 	unsigned long		ns_nrsvsegs;
191 	unsigned long		ns_first_data_block;
192 	int			ns_inode_size;
193 	int			ns_first_ino;
194 	u32			ns_crc_seed;
195 
196 	/* /sys/fs/<nilfs>/<device> */
197 	struct kobject ns_dev_kobj;
198 	struct completion ns_dev_kobj_unregister;
199 	struct nilfs_sysfs_dev_subgroups *ns_dev_subgroups;
200 };
201 
202 #define THE_NILFS_FNS(bit, name)					\
203 static inline void set_nilfs_##name(struct the_nilfs *nilfs)		\
204 {									\
205 	set_bit(THE_NILFS_##bit, &(nilfs)->ns_flags);			\
206 }									\
207 static inline void clear_nilfs_##name(struct the_nilfs *nilfs)		\
208 {									\
209 	clear_bit(THE_NILFS_##bit, &(nilfs)->ns_flags);			\
210 }									\
211 static inline int nilfs_##name(struct the_nilfs *nilfs)			\
212 {									\
213 	return test_bit(THE_NILFS_##bit, &(nilfs)->ns_flags);		\
214 }
215 
216 THE_NILFS_FNS(INIT, init)
217 THE_NILFS_FNS(DISCONTINUED, discontinued)
218 THE_NILFS_FNS(GC_RUNNING, gc_running)
219 THE_NILFS_FNS(SB_DIRTY, sb_dirty)
220 
221 /*
222  * Mount option operations
223  */
224 #define nilfs_clear_opt(nilfs, opt)  \
225 	((nilfs)->ns_mount_opt &= ~NILFS_MOUNT_##opt)
226 #define nilfs_set_opt(nilfs, opt)  \
227 	((nilfs)->ns_mount_opt |= NILFS_MOUNT_##opt)
228 #define nilfs_test_opt(nilfs, opt) ((nilfs)->ns_mount_opt & NILFS_MOUNT_##opt)
229 #define nilfs_write_opt(nilfs, mask, opt)				\
230 	((nilfs)->ns_mount_opt =					\
231 		(((nilfs)->ns_mount_opt & ~NILFS_MOUNT_##mask) |	\
232 		 NILFS_MOUNT_##opt))					\
233 
234 /**
235  * struct nilfs_root - nilfs root object
236  * @cno: checkpoint number
237  * @rb_node: red-black tree node
238  * @count: refcount of this structure
239  * @nilfs: nilfs object
240  * @ifile: inode file
241  * @inodes_count: number of inodes
242  * @blocks_count: number of blocks
243  * @snapshot_kobj: /sys/fs/<nilfs>/<device>/mounted_snapshots/<snapshot>
244  * @snapshot_kobj_unregister: completion state for kernel object
245  */
246 struct nilfs_root {
247 	__u64 cno;
248 	struct rb_node rb_node;
249 
250 	refcount_t count;
251 	struct the_nilfs *nilfs;
252 	struct inode *ifile;
253 
254 	atomic64_t inodes_count;
255 	atomic64_t blocks_count;
256 
257 	/* /sys/fs/<nilfs>/<device>/mounted_snapshots/<snapshot> */
258 	struct kobject snapshot_kobj;
259 	struct completion snapshot_kobj_unregister;
260 };
261 
262 /* Special checkpoint number */
263 #define NILFS_CPTREE_CURRENT_CNO	0
264 
265 /* Minimum interval of periodical update of superblocks (in seconds) */
266 #define NILFS_SB_FREQ		10
267 
268 static inline int nilfs_sb_need_update(struct the_nilfs *nilfs)
269 {
270 	u64 t = get_seconds();
271 
272 	return t < nilfs->ns_sbwtime ||
273 		t > nilfs->ns_sbwtime + nilfs->ns_sb_update_freq;
274 }
275 
276 static inline int nilfs_sb_will_flip(struct the_nilfs *nilfs)
277 {
278 	int flip_bits = nilfs->ns_sbwcount & 0x0FL;
279 
280 	return (flip_bits != 0x08 && flip_bits != 0x0F);
281 }
282 
283 void nilfs_set_last_segment(struct the_nilfs *, sector_t, u64, __u64);
284 struct the_nilfs *alloc_nilfs(struct super_block *sb);
285 void destroy_nilfs(struct the_nilfs *nilfs);
286 int init_nilfs(struct the_nilfs *nilfs, struct super_block *sb, char *data);
287 int load_nilfs(struct the_nilfs *nilfs, struct super_block *sb);
288 unsigned long nilfs_nrsvsegs(struct the_nilfs *nilfs, unsigned long nsegs);
289 void nilfs_set_nsegments(struct the_nilfs *nilfs, unsigned long nsegs);
290 int nilfs_discard_segments(struct the_nilfs *, __u64 *, size_t);
291 int nilfs_count_free_blocks(struct the_nilfs *, sector_t *);
292 struct nilfs_root *nilfs_lookup_root(struct the_nilfs *nilfs, __u64 cno);
293 struct nilfs_root *nilfs_find_or_create_root(struct the_nilfs *nilfs,
294 					     __u64 cno);
295 void nilfs_put_root(struct nilfs_root *root);
296 int nilfs_near_disk_full(struct the_nilfs *);
297 void nilfs_fall_back_super_block(struct the_nilfs *);
298 void nilfs_swap_super_block(struct the_nilfs *);
299 
300 
301 static inline void nilfs_get_root(struct nilfs_root *root)
302 {
303 	refcount_inc(&root->count);
304 }
305 
306 static inline int nilfs_valid_fs(struct the_nilfs *nilfs)
307 {
308 	unsigned int valid_fs;
309 
310 	down_read(&nilfs->ns_sem);
311 	valid_fs = (nilfs->ns_mount_state & NILFS_VALID_FS);
312 	up_read(&nilfs->ns_sem);
313 	return valid_fs;
314 }
315 
316 static inline void
317 nilfs_get_segment_range(struct the_nilfs *nilfs, __u64 segnum,
318 			sector_t *seg_start, sector_t *seg_end)
319 {
320 	*seg_start = (sector_t)nilfs->ns_blocks_per_segment * segnum;
321 	*seg_end = *seg_start + nilfs->ns_blocks_per_segment - 1;
322 	if (segnum == 0)
323 		*seg_start = nilfs->ns_first_data_block;
324 }
325 
326 static inline sector_t
327 nilfs_get_segment_start_blocknr(struct the_nilfs *nilfs, __u64 segnum)
328 {
329 	return (segnum == 0) ? nilfs->ns_first_data_block :
330 		(sector_t)nilfs->ns_blocks_per_segment * segnum;
331 }
332 
333 static inline __u64
334 nilfs_get_segnum_of_block(struct the_nilfs *nilfs, sector_t blocknr)
335 {
336 	sector_t segnum = blocknr;
337 
338 	sector_div(segnum, nilfs->ns_blocks_per_segment);
339 	return segnum;
340 }
341 
342 static inline void
343 nilfs_terminate_segment(struct the_nilfs *nilfs, sector_t seg_start,
344 			sector_t seg_end)
345 {
346 	/* terminate the current full segment (used in case of I/O-error) */
347 	nilfs->ns_pseg_offset = seg_end - seg_start + 1;
348 }
349 
350 static inline void nilfs_shift_to_next_segment(struct the_nilfs *nilfs)
351 {
352 	/* move forward with a full segment */
353 	nilfs->ns_segnum = nilfs->ns_nextnum;
354 	nilfs->ns_pseg_offset = 0;
355 	nilfs->ns_seg_seq++;
356 }
357 
358 static inline __u64 nilfs_last_cno(struct the_nilfs *nilfs)
359 {
360 	__u64 cno;
361 
362 	spin_lock(&nilfs->ns_last_segment_lock);
363 	cno = nilfs->ns_last_cno;
364 	spin_unlock(&nilfs->ns_last_segment_lock);
365 	return cno;
366 }
367 
368 static inline int nilfs_segment_is_active(struct the_nilfs *nilfs, __u64 n)
369 {
370 	return n == nilfs->ns_segnum || n == nilfs->ns_nextnum;
371 }
372 
373 static inline int nilfs_flush_device(struct the_nilfs *nilfs)
374 {
375 	int err;
376 
377 	if (!nilfs_test_opt(nilfs, BARRIER) || nilfs->ns_flushed_device)
378 		return 0;
379 
380 	nilfs->ns_flushed_device = 1;
381 	/*
382 	 * the store to ns_flushed_device must not be reordered after
383 	 * blkdev_issue_flush().
384 	 */
385 	smp_wmb();
386 
387 	err = blkdev_issue_flush(nilfs->ns_bdev, GFP_KERNEL, NULL);
388 	if (err != -EIO)
389 		err = 0;
390 	return err;
391 }
392 
393 #endif /* _THE_NILFS_H */
394